10 Úno
2022
10 Úno
'22
10:57
Hi,
I've got a staging environment running, with two signers signing a
staging version of the .is zone (~115k records).
The staging servers are configured with 2 GB RAM and 4 CPU cores,
running on FreeBSD 12.2.
We've experienced that knot crashes because the server runs out of
memory. The zone is configured with:
zonefile-sync: -1
zonefile-load: difference-no-serial
journal-content: all
One of the signers, after an hour of running is showing 215 MB resident
memory used.
The other signer, that's been running for a whole day is showing 1582 MB
resident memory.
I attached a screenshot showing the amount of memory used over a period
of time, and the graph shows that the amount of memory used very
suddenly increases.
I assume the servers are using a lot of memory for the journals, but I'd
like to understand why the sudden increase in used memory, and what to
expect about needed memory?
.einar
10 Úno
10 Úno
11:07
Hi Einar,
Memory consumption is a difficult topic. Could you try setting `server.background-workers:
1`? In some cases it helps.
I don't think the problem is in the journal.
BTW, we did some memory optimizations in the development version (3.2). Do you instal Knot
from ports or as a binary package?
We can provide a tarball for testing.
Best,
Daniel
On 2/10/22 10:57, Einar Bjarni Halldórsson wrote:
Hi,
I've got a staging environment running, with two signers signing a staging version of
the .is zone (~115k records).
The staging servers are configured with 2 GB RAM and 4 CPU cores, running on FreeBSD
12.2.
We've experienced that knot crashes because the server runs out of memory. The zone
is configured with:
zonefile-sync: -1
zonefile-load: difference-no-serial
journal-content: all
One of the signers, after an hour of running is showing 215 MB resident memory used.
The other signer, that's been running for a whole day is showing 1582 MB resident
memory.
I attached a screenshot showing the amount of memory used over a period of time, and the
graph shows that the amount of memory used very suddenly increases.
I assume the servers are using a lot of memory for the journals, but I'd like to
understand why the sudden increase in used memory, and what to expect about needed
memory?
.einar
--
12:01
Hi Daniel,
On 10.2.2022 10:07, Daniel Salzman wrote:
Memory consumption is a difficult topic. Could you try setting
`server.background-workers: 1`? In some cases it helps.
I don't think the problem is in the journal.
I'll try setting server.background-workers to 1.
BTW, we did some memory optimizations in the
development version (3.2).
Do you instal Knot from ports or as a binary package?
We can provide a tarball for testing.
We have a poudriere server that builds binary packages for all servers.
We could add a custom port to build 3.2, or just build it locally on the
servers to test first.
I've bumped the staging servers to 4 GB RAM, I want to see if the memory
usage tops out.
.einar
14:12
Hi Einar,
from the graph it seems that the memory consumption rocketed in an
instant manner, not in a progressive manner. Am I looking correctly?
Could you please read out from logs what was Knot doing around that time?
Thanks,
Libor
BTW, we did some memory optimizations in the
development version
(3.2). Do you instal Knot from ports or as a binary package?
We can provide a tarball for testing.
We have a poudriere server that builds binary packages for all servers.
We could add a custom port to build 3.2, or just build it locally on
the servers to test first.
I've bumped the staging servers to 4 GB RAM, I want to see if the
memory usage tops out.
.einar
--
828
days inactive
828
days old
3 comments
3 participants
participants (3)
-
Daniel Salzman -
Einar Bjarni Halldórsson -
libor.peltan